Model performance optimization method, device, equipment and storage medium

Abstract

The invention discloses a model performance optimization method, device, equipment and storage medium. The method includes: inputting preset negative example training data into the model to be trained to obtain a prediction result; when the prediction result represents the negative example training data is predicted as When the probability of a negative example is greater than the preset threshold, correct the loss function value corresponding to the negative example training data, wherein the corrected loss function value is smaller than the pre-corrected loss function value; adjust the model to be trained according to the corrected loss function value The model parameters of the to-be-trained model are trained to obtain the target model. The present invention reduces the influence of wrongly labeled negative example training data on model training, thereby improving the prediction or classification accuracy of the target model and improving model performance.

Description

technical field [0001] The present invention relates to the technical field of machine learning, in particular to a model performance optimization method, device, equipment and storage medium. Background technique [0002] With the rapid development of machine learning technology, the scope of machine learning applications is becoming wider and wider. The training optimization of machine learning models has also become the direction of continuous exploration of the entire industry. In the training process of existing machine learning models, positive and negative examples are used to train the model. For example, when training an image target detection model, images containing targets (positive examples) and images not containing targets are used. (Negative example) for training. However, in many scenarios, there will be cases where they are actually positive examples, but they are wrongly marked as negative examples due to various reasons, and then they are treated as nega...

AI Technical Summary

Problems solved by technology

[0003] The main purpose of the present invention is to provide a model performance optimization method, device, equipment, and storage medium, aiming to solve the problem of inaccurate prediction or classification of the model obtained by training due to the use of wrongly labeled negative examples to train the model. question

Images